Random and Systematic Error in Voting in Presidential Elections

• Date: 01 September 2013
• DOI: 10.1177/1065912912459565
• Author: Sean Richey
• Published date: 01 September 2013
• Subject Matter: Articles

459565PRQXXX10.1177/10659129124
59565Political Research QuarterlyRichey
Regular Article
Political Research Quarterly
66(3) 645 –657
Random and Systematic Error in
© 2012 University of Utah
Reprints and permissions:
sagepub.com/journalsPermissions.nav
Voting in Presidential Elections
DOI: 10.1177/1065912912459565
prq.sagepub.com
Sean Richey1
Abstract
Condorcet’s theory of voting rests on the crucial proposition that voting errors are random and not systematic. Using
Lau and Redlawsk’s voting correctly measure, I test whether voting error is systematic or random in presidential
elections from 1972 to 2004. I show that errors are systematically skewed toward Republican candidates. I also show
that the level of skew of incorrect voting has led to the incorrect candidate being elected in three out the last nine
elections. In addition, I find that greater skew in Republican campaign spending increases skew of incorrect votes
toward Republican candidates.
Keywords
campaign finance, voting correctly, Condorcet’s jury theorem
Do irrational vote choices aggregate in a way that cancels
Up to this point, however, testing the theory has been dif-
out the mistakes? The theory of aggregation is one of the
ficult because it has been hard to determine whether vot-
most important explanations of why democracy succeeds
ers are rational.
even though citizens are often uninformed, apathetic, and
To test the aggregation theory, I use Lau and Redlawsk’s
disengaged from politics (Page and Shapiro 1992). This
voting correctly data. Lau and Redlawsk (1997, 2006)
theory suggests that elections are “as-if” rational, because
develop a measurement system to determine whether vot-
voting errors are random, and therefore equally distrib-
ers choose the candidate that is in accordance with their
uted, which will not affect most outcomes (Fey 2003;
views on politics. This tests whether voters are consistent
Lindner 2005). If errors are randomly distributed, it is
in their preferences and choices, which is a minimal defini-
easy to show how elections can lead to optimal outcomes
tion of rationality that is used by many theorists (e.g.,
even with a large number of erroneous votes (List and
Arrow 1951). Up to this point, researchers have focused on
Goodin 2001). But, what if the errors are systematic due
the total number of correct votes (e.g., Baum and Jamison
to factors such as campaign manipulation of voters? Vot-
2006). Although extremely interesting, the prior literature
ers can be manipulated by fearmongering, race-baiting,
on correct voting has not used this measure to test the
and so on (Brader 2006; Mendelberg 2001). If some cam-
aggregation theory. Using this measure developed from
paigns are more manipulative, then the errors will have a
survey data from the American National Election Studies
skewed distribution, and voting will not produce random
(ANES) from presidential elections from 1972 to 2004, I
errors of equal balance and size. If the errors are skewed,
show below that voting errors systematically skew to
the aggregation theory will not hold, and elections may
Republican candidates. Republican candidates are able to
produce suboptimal results. Thus, we need to determine
gain about 4.6 percentage points more incorrect votes than
the presence and size of skew in the aggregation of voting
Democratic candidates in that time, but with considerable
errors to test this canonical theory.
variation between elections. I also show that the level of
The miracle of aggregation is commonly used to
skew in incorrect voting has led to the incorrect candidate
defend democracy (e.g., List and Goodin 2001). If it turns
being elected in three1 out the last nine elections, and in
out to be unfounded, or only partially successful, it will
two of the three mistaken outcomes the irrational voting
inform us on the potential for democracy to reach supe-
rior outcomes. With the spread of democracy to nearly all
1Georgia State University, Atlanta, USA
parts of the globe, it is crucial to test one of its founda-
tional presuppositions. In addition, by discovering what
Corresponding Author:
Sean Richey, Department of Political Science, Georgia State University,
influences irrational voting for one party but not the other,
38 Peachtree Center Ave., Suite 1005, Atlanta, GA 30303-2514, USA.
it may point toward certain reforms that can be beneficial.
Email: srichey@gsu.edu

---

646
Political Research Quarterly 66(3)
favored a Republican candidate. In addition, I use a multi-
that reach a greater number of voters, and some of the mes-
nomial probit regression model of National Election
sages that they deliver will be manipulative. Campaigns
Studies (NES) data from the same elections and include
need to persuade voters. Based on a massive literature on
campaign spending data to determine the different influ-
persuasion in social psychology, we know that persuasion
ences on Republican and Democratic incorrect voting. I
can be direct or peripheral (see Petty, Cacioppo, and
show that although political sophistication lowers incorrect
Schumann 1983). The direct route uses rational consider-
voting for both sides, skew in campaign spending influ-
ations such as issue voting to gain voters’ support, whereas
ences incorrect voting for Republican candidates.
the peripheral route uses nonrational cues and misleading
This research proceeds as follows. First, I discuss the
emotional appeals to gain compliance. If one campaign has
theory of aggregation, while considering why small devi-
more resources that allow it to be more effective in periph-
ations from equal aggregation will be important. Then I
eral route persuasion, the errors will be systematic.
review the data sources and methods of analysis. Finally,
Manipulation is particularly necessary for campaigns that
I examine voting incorrectly for each party and conclude
are in the numerical minority, because they will need to get
by discussing the results.
some voters to vote incorrectly to win. As Page and Shapiro
(1992, 26) say, “Just as juries can be moved by deceptive
Aggregation and Manipulation
advocates or by perjured testimony, the public can be
deceived by propaganda.” In addition, we know that cam-
The miracle of aggregation derives from Condorcet’s
paigns often offer misleading information to bolster their
analysis of jury voting (Young 1988).2 It shows that ran-
cause (Callander and Wilkie 2007). Thus, a useful starting
domly distributed erroneous votes will balance each
assumption is that campaigns will try to get voters to vote for
other out, leaving a reasonable outcome (Austen-Smith
candidates whom they would not prefer under fully rational
and Banks 1996). Thus, if ten out of twelve jurors make
decision making. In sum, due to skew in campaign resources
random errors, they should split five and five, and the
and the need for campaigns to manipulate voters, I will test
outcome will be decided by the two who vote correctly.
below whether skew in resources leads to skew in errors.
The law of large numbers shows that this as-if rational
Second, in addition to manipulation, many voters may
outcome will be almost certain in any large election if the
have difficulty because of limited interest in the cam-
errors are randomly distributed (Ladha 1992). Based on
paign and the general difficulty in processing information
this analysis, it is possible that elections or public opinion
that may be statistically correlated with different social
can have rational outcomes generally, even with a large
groups that support different parties (Berinsky 1999). For
level of irrationality (Page and Shapiro 1992, 34).
example, say those with lower education disproportion-
Erikson, MacKuen, and Stimson (2002, 429) state that a
ately should correctly vote for Democratic candidates,
major reason for democracy’s success is the “reduction
but because of their lower education, they make more
of noise due to the miracle of aggregation.”
mistakes. If those with higher education vote correctly
But underlying this logic is the crucial assumption that
more often and disproportionately support Republican
the errors are random (Kaplan 2007). If there are nonran-
candidates, then the theory of aggregation will not hold.
dom errors, then this assumption does not hold, and the
The errors will be systematically skewed based on educa-
outcome will be biased (Berg 1996). In a review of this
tion level.3 This is troublesome for democratic theory,
literature, Kuklinski and Quirk (2000, 160) note that “it is
because it suggests that the party associated with higher
far from clear that the errors in individual citizen’s political
education levels will be able to gain more incorrect votes
judgments are largely random.” Althaus (2003) conducts a
consistently.4 In addition, then, to skew in campaign
thorough in-depth analysis of public opinion aggregation,
resources, skew in education between the parties’ sup-
and finds that there are several reasons for aggregation to
porters will also lead to skew in voting errors and thereby
not be random or equally distributed, in addition to other
thwart Condorcet’s logic.
problems such as skew in response bias to surveys.
Generally, voters do not have the knowledge, skills, or
Fournier (2006) also shows that aggregated intended vote
time to overcome manipulation. We know there is a large
choices differ from a simulated...